Text correction in distorted label images by applying biquadratic transformation

Natural scene images often contain structural deformations or distortions. Especially label images are wrapped around cylindrical objects, thus when viewed in a skewed direction, the text on the label appear severely distorted. Because of the distorted text, it makes some difficult to recognize the text. Recognition of such distorted text on label images can be a source of incorrect recognition of the label content. In this paper, we present two methods on the correction of the distorted text, one by using mapping function and another by the biquadratic transformation function. First, boundary lines of the label are detected by Hough transform and Bezier curve approximation. Experimental results show the proposed methods correctly restore the original rectangular shape of the label area.